[Rspamd-Users] Whitelisting recipients and bypass bayes autolearning

[Andy Balholm]

The autolearn setting can be a custom Lua function. I would assume it's 
possible to do what you want that way. But the documentation is rather 
sparse.

Andy

On 10/13/21 4:32 AM, Aban Dokht wrote:
> Hello,
>
>
> I need to formward SPAM, phishing and other suspect emails to a 
> external security company, let's say samples at example.com.
>
> To not block outgouing mail, I have added the following to 
> /etc/rspamd/local.d/multimap.conf:
>
> LOCAL_WL_RECIPIENT {
>     type = "rcpt";
>     map = "$LOCAL_CONFDIR/local.d/local_wl_rcpt.map.inc";
>     description = "Recipient is trusted";
>     action= "no action"
>     prefilter=true
> }
>
> and added samples at example.xom to local_wl_rcpt.map.inc.
>
> Mails sent to samples at example.com pass rspamd as expected, score is 0.
>
>
> But emails do get lerned by Bayes, as I have configured Bayes to learn 
> < 0.5
>
> /etc/rspamd/local.d/classifier-bayes.conf:
> autolearn = [0.5, 5]
>
>
> How can I prevent rspamd to learn emails to recipielnt in 
> local_wl_rcpt.map.inc?
>
> Do I really need to set the score > 0.5 in multimap.conf?
>
> Or is there some better way, like action= "no action + no autolearn" ?
>
>
>
> Regards
> Aban
>